Permafrost-carbon mobilization in Beringia caused by deglacial meltwater runoff, sea-level rise and warming

Pre-depositional ages of the alkanes and fatty acids deposited during the LGM at the three core location are >16 000 years (figures 2(i)–(k), table 1). Nowadays, leaf-wax lipids found off Arctic drainage basins underlain by continuous permafrost as well as lipids deposited off Siberian coastlines that currently undergo massive erosion are several thousand years old due to substantial contributions of ancient organic matter from Yedoma (∼5500–∼13 000 years; Lena, Kolyma, Indigirka Rivers (Feng et al 2013, Gustafsson et al 2011); up to ∼18 000 years off eroded coasts (Vonk et al 2014)). By contrast, lipids deposited off permafrost-free catchments are much younger (∼900–4400 years) (Kusch et al 2010). Thus, the relatively high LGM pre-depositional ages in our cores may indicate large proportions of old organic carbon from ancient Yedoma. This interpretation is in line with relatively high values in fββ, CPI_alk and CPI_FA (figures 3(b), (d) and (h)) which render old but little degraded organic matter the dominant source for the alkanes and fatty acids at this time. Our CPI_alk and CPI_FA values (∼4–8) are in the range of CPI-values reported in Siberian and Alaskan Yedoma deposits (3–13) (Sánchez-García et al 2014, Jongejans et al 2018).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Minimal mass accumulation rates of terrigenous biomarkers at all sites (figures 2(i)–(k)) point to relatively little coastal erosion and fluvial discharge of permafrost carbon during the LGM. This is in line with the presence of intact continuous permafrost in Beringia (Lindgren et al 2015), cold sea surface temperatures (SST) in the subarctic North Pacific (Meyer et al 2016, Méheust et al 2018, Maier et al 2018) and a continental climate with very cold winters (Elias 2001, Anderson and Lozhkin 2015).

The early deglaciation is characterized by dissimilar patterns in mass accumulation rates of the different biomarkers. All cores record enhanced burial of old alkanes (pre-depositional age: >11 600 years) between ∼16.8 and 15 kyr BP coincident with Heinrich Stadial 1 (figures 2(i)–(k), 3(b), (d) and (h)). Off the Bering Shelf (cores SO202-18-3/SO202-18-6) this increase is accompanied by enhanced burial of old fatty acids (pre-depositional age: >9600 years) and brGDGT while off the Apuka River (core SO201-2-114KL) and Kamchatka (core SO201-2-12KL) the fatty acids do not follow the increase reported in the alkanes but remain constant until ∼14.6 kyr BP (figures 2(i)–(k), 3(a), (d) and (g)). Distinct minima in CPI_alk (∼3–4) occur concurrently with the enhanced accumulation of alkanes at all sites (figures 3(b), (e) and (h)). These minima indicate that the amplification of alkane export is associated with a drastic increase in the degree of degradation of the alkanes. By contrast, all cores display constant CPI_FA-values across the LGM and early deglaciation (figures 3(b), (e) and (h)) indicating that the degradation state of the fatty acids remains unaltered. These discrepancies in mass-accumulation rates and CPI-values of alkanes and fatty acids imply that the compounds derive from different sources and record different erosion and transport processes.

Possible explanations for the lowering of CPI_alk include biogeochemical degradation of alkanes during land-ocean transport (e.g. Bröder et al 2016), changes in the composition of the vegetation (Bush and McInerney 2013), or contributions of petrogenic organic matter (Bray and Evans 1961). At all cores, the minima in CPI_alk are congruent with pronounced minima in fββ which decreases by 0.3–0.6 units reaching values as low as 0.2–0.4 (figures 3(c), (f) and (i)). These minima in fββ hint at massive reworking of petrogenic organic matter since the lowering of fββ stems from a distinct increase in the diagenetic C₃₁αβS and C₃₁αβR epimers (figure S1) which form under elevated temperatures during burial (van Duin et al 1997, Farrimond et al 1998) and dominate over the biologic C₃₁ββR isomer in our cores (figure S1). Given the coincidence of the minima in CPI_alk and fββ it is most likely that the low CPI_alk results from significant contributions of petrogenic alkanes. As the CPI_alk values (3–4) still exceed the characteristic value of petrogenic organic matter (∼1) (Bray and Evans 1961), the old alkanes (pre-depositional age of 11 600–19 400 years) deposited at the core locations during the early deglaciation probably derive from a mixture of fossil petrogenic organic matter and old but less degraded material, most likely derived from Yedoma.

In contrast to the alkanes, the source of the fatty acids likely did not change significantly after the LGM considering that the CPI_FA is relatively constant across the LGM and the early deglaciation. The high pre-depositional ages of the fatty acids (9600–13 700 years, figures 2(i)–(k)) point to Yedoma deposits as the predominant source, during the LGM as well as the early deglaciation.

As such, the concurrent increase in fatty acids and brGDGT off the Bering Shelf (cores SO202-18-3/SO202-18-6; figures 2(i)–(k) and 3(a)) points to an initial destabilization of permafrost soils and subsequent organic matter remobilization in Alaska and on the Bering Land Bridge. By contrast, low export-rates of old fatty acids from the Apuka catchment and Kamchatka (cores SO201-2-114KL and SO201-2-12KL) indicate that permafrost likely remained stable in these regions throughout the early deglaciation until ∼14.6 kyr BP (figures 2(j) and (k)).

The onset of increased accumulation of old terrigenous biomarkers off the Bering Shelf (cores SO202-18-3/SO202-18-6; figures 2(i) and 3(a)) between ∼16.8 and 15 kyr BP coincides with the beginning retreat of the Cordilleran and Laurentide ice sheets and Alaskan mountain glaciers at ∼19–16 kyr BP (Dyke 2004, Badding et al 2013, Maier et al 2018). A related massive meltwater input into the Yukon River system (Duk-Rodkin and Hughes 1994, Kennedy et al 2010), might have caused severe riverbank erosion of the organic-rich Yedoma deposits in Alaska and on the Bering Land Bridge, subsequently enhancing the accumulation of permafrost-derived organic matter off the Bering Shelf (cores SO202-18-3/SO202-18-6). Also, the enhanced burial of petrogenic carbon at all sites matches the deglaciation of the Brooks Range and the final drainage of the proglacial Lake Old Crow (∼16.8–15 kyr BP; figures 1 and 3) (Zazula et al 2004, Badding et al 2013). As mature deposits have been found at both locations (Barnes 1967, Alaska State Office 2009), it is likely that the massive reworking of petrogenic carbon is associated with these local deglaciation events in Alaska. The simultaneous deposition of petrogenic carbon along the core transect (figures 3(b), (e) and (h)) suggests that the petrogenic fraction of the Yukon load spilled out into the Northwest Pacific, and was likely transported by the counterclockwise surface circulation of the Bering Sea (figure 1). A profound influence from the Yukon load on the fatty acids deposited off the Apuka catchment and Kamchatka (cores SO201-2-114KL and SO201-2-12KL) seems unlikely considering that cores SO201-2-114KL and SO201-2-12KL display a different pattern in mass accumulation rates of fatty acids than cores SO202-18-3/SO202-18-6 (figures 2(i)–(k) and 3(a)).

The increased accumulation of fatty acids and brGDGTs off the Bering Shelf (cores SO202-18-3/SO202-18-6) is also synchronous with a first maximum in the global rate of sea-level rise (Lambeck et al 2014) (figure 2). The synchrony implies that rapid sea-level rise may have eroded permafrost deposits on the Bering Shelf during this interval (figures 2(d) and (i)).

Thawing of permafrost in the hinterland (Lindgren et al 2018) would be another mechanism increasing the fluvial export of old permafrost carbon from the Yukon and Anadyr drainage basins by active layer deepening, thaw-slumps along riverbanks and intensified overland drainage. This view is in line with rising summer insolation (figure 2(d)) (Berger and Loutre 1991) and air temperature in some parts of interior Alaska at this time (Kurek et al 2009). However, cold SST in the Bering Sea and the North Pacific (Meyer et al 2016, Méheust et al 2018) suggest that at least in Pacific Beringia (i.e. the Apuka catchment, Kamchatka, and the southern Bering Land Bridge) low air temperatures may have prevented permafrost from thawing. This is corroborated by the low mass accumulation rates of the fatty acids off the Apuka River mouth and Kamchatka (cores SO201-2-114KL and SO201-2-12KL, figures 2(j) and (k)). Since only very few Beringian thaw lakes date back to the early deglaciation (Walter et al 2007, Walter Anthony et al 2014) (figure 2(f)), thermokarst processes were probably limited in Siberia and Alaska pointing to mostly intact permafrost conditions across Beringia. P_aq shows a first maximum in wetland expansion at that time (figure 2(h)) congruent with the intensification of organic matter accumulation off the Bering Shelf (cores SO202-18-3/SO202-18-6; figure 2(i)). This may hint towards beginning ecosystem changes in the permafrost landscape of central/eastern Beringia associated with wetter conditions possibly due to the massive meltwater floods. However, in light of the substantial petrogenic contributions to the alkanes at that time, the P_aq—which is built on the relative abundance of alkane homologs (see supplementary material)—probably reflects altered relative abundances of alkane homologs that are associated with the petrogenic portion rather than with changing vegetation. Thus, widespread permafrost decomposition and thaw-induced riverine transport of permafrost carbon during the early deglaciation seem unlikely.

Around ∼15 kyr BP prominent increases in fββ and CPI_alk (figures 3(b), (c), (e), (f), (h) and (i)) mark the cessation of major petrogenic contributions at all sites. Relatively high values in fββ, CPI_alk, CPI_FA throughout the late deglaciation and the Holocene (figures 3(b), (c), (e), (f), (h) and (i)) together with high pre-depositional ages (9600–12 300 years; figures 2(i)–(k); table 1) suggest that alkanes and fatty acids dominantly derive from old but less degraded deposits, most likely permafrost soils.

Our records display two major intervals of enhanced inflow of permafrost carbon, that are consistent with the Bølling-Allerød and Preboreal interstadials (figures 2(c), (i)–(k)). Both intervals are found in mass accumulation rates of terrestrial biomarkers off the Bering Shelf (cores SO202-18-3/SO202-18-6) and off Kamchatka (site SO201-2-12KL), while in core SO201-2-114KL only one peak during the Preboreal is detected (figure 2(k)). One may speculate that permafrost remained intact in the small Apuka catchment and adjacent shelves until ∼11.5 kyr BP.

The mass accumulation rates rapidly culminate in two maxima at ∼14.6 and ∼11.5 kyr BP (figures 2(i)–(k)). These coincide with mwp-1A and mwp-1B, (figure 2(d)) suggesting that rapid flooding of the Bering Land Bridge and the Pacific shelves adjacent to Siberia and Kamchatka caused massive coastal erosion of permafrost deposits. This is in accordance with the flooding scenario of the Bering Shelf (Manley 2002) which suggests that most of the shelf was flooded between 15 and 10 kyr BP, consistent with high rates of sea-level rise (Lambeck et al 2014).

The maxima of massive translocation of permafrost carbon are also synchronous with rapid warming in the North-Pacific realm (Kurek et al 2009, Meyer et al 2016, 2017, Méheust et al 2018) during the onsets of the Bølling-Allerød and Preboreal interstadials (NGRIP 2004) at 14.6 and 11.5 kyr BP. Together with elevated summer insolation (Berger and Loutre 1991) warming may have caused widespread permafrost thaw in the Kamchatka, Apuka, Anadyr and Yukon River watersheds. Subsequent fluvial discharge of the mobilized organic matter may have increased the accumulation of terrigenous biomarkers at our sites. Walter Anthony et al (2014) report pronounced expansion of thermokarst lakes in Siberia and Alaska during the interstadials which points to widespread abrupt permafrost thaw across Beringia. Concurrent wetland development in the Anadyr and Yukon catchments—as reported by the increased P_aq values (cores SO202-18-3/SO202-18-6, figure 2(h))—suggests generally wetter conditions and possibly amplified overland drainage in Siberia and Alaska. However, the thermokarst-lake expansion peaked during the second half of the warm-spells (figure 2(f)) lagging the abrupt warming and melt-water pulses at 14.6 and 11.5 kyr BP (Reyes and Cooke 2011, Walter Anthony et al 2014) as well as the rapid amplification of biomarker accumulation at our sites (figure 2(d), (i)–(k)) at this time. Thus, widespread thaw-induced mobilization of permafrost carbon and subsequent fluvial transport most likely initiated after the melt-water pulses implying that the maxima in accumulation of permafrost carbon at our sites around 14.6 and 11.5 kyr BP (figures 2(i)–(k)) are probably primarily controlled by coastal erosion associated with mwp-1A and mwp-1B. Thaw-induced mobilization of permafrost carbon in the drainage basins of the rivers likely contributed to the elevated mass-accumulation rates of the biomarkers shortly after the melt-water pulses.

Additionally, riverbank erosion of permafrost due to retreating American ice sheets (Keigwin et al 2018) and mountain glaciers on Kamchatka (Bigg et al 2008) may have proceeded throughout the late deglaciation.

For the Younger Dryas (12.9-11.5 kyr BP), all cores indicate a slackening of the accumulation of terrigenous biomarkers (figures 2(i)–(k)) which coincides with reduced thaw lake development (Walter Anthony et al 2014), wetland vegetation (lowered P_aq, cores SO202-18-3/SO202-18-6) and rates of sea-level change. Thus, permafrost decomposition in the interior and along the coast likely weakened, probably in response to cooling in the North Pacific realm (Meyer et al 2016, Méheust et al 2018) and decelerated sea-level rise (Lambeck et al 2014).

All cores display highest mass accumulation rates during the Preboreal indicating that the remobilization of permafrost carbon reached its maximum at this time (figures 2(i)–(k)). Coastal erosion of permafrost may have become maximal during mwp-1B since the flooding scenario for the Bering Shelf shows that mwp-1B inundated a larger area than mwp-1A (Manley 2002). Thermokarst-lake expansion (Walter Anthony et al 2014) and wetland extent (P_aq; cores SO202-18-3/SO202-18-6, figures 2(f) and (h)) became maximal (Reyes and Cooke 2011). This observation likely attests to maximal permafrost thaw in Siberia and Alaska considering that during the Preboreal summer insolation and regional temperatures reached the highest values since the LGM (Berger and Loutre 1991, Kurek et al 2009, Meyer et al 2016).

After the Preboreal (∼9–10 kyr BP), all cores display a distinct decrease and subsequent constancy of the mass-accumulation rates of terrigenous biomarkers (figures 2(i)–(k)) suggesting that the remobilization of old permafrost-derived organic carbon diminished and stabilized. Coastal erosion rates likely decreased as the inundation of the Bering Shelf was almost completed around 10 kyr BP (Manley 2002) after the opening of the Bering Strait (∼11 kyr BP; Jakobsson et al 2017) and mwp-1B (Lambeck et al 2014). Thermokarst lake (Walter Anthony et al 2014) and wetland formation (P_aq cores SO202-18-3/SO202-18-6; figure 2(h)) weakened and stabilized after the Preboreal potentially marking a thorough reduction of inland permafrost thaw and the related discharge of permafrost carbon to our sites (figures 2(i)–(k)).

During the late Holocene, the pre-depositional ages found off Kamchatka (∼4500 years; core-top core SO201-2-12KL; figure 2, table 1) are significantly younger than during the deglaciation and the LGM (>6000 years; figure 2(k), table 1). The decrease in age is probably due to the rather limited permafrost extent on Kamchatka at present as most of the peninsula is permafrost free with only a few patches of sporadic and alpine permafrost (Brown et al 1998).

We identify two major intervals of massive land-ocean transfer of old permafrost carbon across Beringia's Pacific coast that occur during the Bølling-Allerød and Preboreal warm-spells. We find that these two intervals initiated rapidly in response to abrupt sea-level rise during mwp-1A and mwp-1B and likely received contributions from thaw-induced processes in the hinterland shortly after the melt-water pulses. Our mass-accumulation rates indirectly mark intervals of enhanced release of ancient CO₂ from degrading permafrost (Winterfeld et al 2018) as exposure of previously freeze-locked carbon to microbial activity in the rivers, the marine water column and at the sediment-water interface of the ocean might have resulted in its rapid decomposition.

As such, the centennial-scale amplification of permafrost erosion during mwp-1A and mwp-1B is probably indicative of similarly rapid CO₂ emissions into the atmosphere. Within the uncertainties of the chronologies (< ±300 years) (Max et al 2012, Kuehn et al 2014) this rapid carbon release is synchronous with the rapid rises in atmospheric CO₂ at 14.6 and 11.5 kyr BP (Marcott et al 2014) (figure 2(b)). The same chronology of processes causing permafrost decomposition as found in our study (coastal erosion followed by inland permafrost thaw and river discharge) has also been inferred from biomarker records from the Okhotsk Sea off the Amur River mouth (Winterfeld et al 2018). The consistency implies that at least in the North-Pacific realm sea-level rise was a major control for abrupt remobilization of permafrost carbon at 14.6 and 11.5 kyr BP. As for the Arctic sector of Beringia, deglacial reworking of old permafrost carbon caused by inundation of the Chukchi Shelf (at ∼13 kyr BP, figure 2(f)) has been recently reported (Martens et al 2019). Yet, it remains unresolved whether Beringia's Arctic shelves (East Siberian Arctic and Chukchi Shelves, figure 1) released carbon at 14.6 and 11.5 kyr BP along with the Pacific shelves (Martens et al 2019). Tesi et al (2016) document massive fluvial discharge of permafrost carbon from the Lena River into the Laptev Sea around 11.5 kyr BP suggesting that thaw-induced degradation of permafrost in drainage basin of the Lena coincided with rapidly rising atmospheric CO₂ levels (figure 2(f)). Altogether these data confirm model-based hypothesis according to which carbon release from coastal as well as inland permafrost decomposition contributed substantially to the rapid deglacial rises in atmospheric CO₂ (Köhler et al 2014, Winterfeld et al 2018).

Extrapolating biomarker data from the small Okhotsk-Sea shelves to the extensive Bering, Chukchi and East Siberian Arctic Shelves Winterfeld et al (2018) estimated with a carbon cycle model that flooding of this area alone may have accounted for about half of the abrupt rises in atmospheric CO₂ (∼6 ppm) at 14.6 and 11.5 kyr BP thereby causing a drop of 6–8‰ in atmospheric Δ¹⁴C. Our data reveal that the timing of carbon mobilization (peaks in mass-accumulation rates, figures 2(i), (j) and (l)) as well as the pre-depositional ages agree well between the Sea of Okhotsk (∼5000–10 000 years; figure 2(l)) and the Bering Sea (∼6900–10 500 years; figure 2(i), table 1, cores SO202-18-3/SO202-18-6) at these times. Given these consistencies our findings underpin the results of the carbon cycle simulation by Winterfeld et al (2018) highlighting the potential key role of Beringia in the deglacial permafrost-carbon feedback, particularly at 14.6 and 11.5 kyr BP.

At 16.5 kyr BP, flooding of the Bering, Chukchi and East Siberian Arctic Shelves may only explain up to a fourth of the observed sudden rise in atmospheric CO₂ according to Winterfeld et al (2018). Our data suggest that next to incipient inundation of the Bering Shelf massive meltwater discharge due to glacier retreat caused riverbank erosion of permafrost soils in the Yukon catchment and concomitantly mobilized fossil, petrogenic carbon around that time (figures 3(a), (d) and (g)). This process could have additionally fueled the observed rise in atmospheric CO₂ at 16.5 kyr BP, if the mobilized material was at least partly remineralized during transport processes prior to final burial in marine sediments and if the melt-water induced mobilization of old carbon was a broad-scale phenomenon around the large Eurasian and American ice sheets.